CN105303151B - The detection method and device of human face similarity degree - Google Patents
The detection method and device of human face similarity degree Download PDFInfo
- Publication number
- CN105303151B CN105303151B CN201410306792.5A CN201410306792A CN105303151B CN 105303151 B CN105303151 B CN 105303151B CN 201410306792 A CN201410306792 A CN 201410306792A CN 105303151 B CN105303151 B CN 105303151B
- Authority
- CN
- China
- Prior art keywords
- face
- pixel data
- picture
- pixel
- face contour
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Landscapes
- Image Analysis (AREA)
Abstract
The invention discloses a kind of detection methods of human face similarity degree, this method comprises: obtaining the face information in detection picture, face information is stored as face contour pixel data corresponding to face;Face contour pixel data are read according to preset read direction, face contour pixel data and the difference value of face contour pixel data corresponding in reference picture read is calculated by color sub-pixels;According to the difference value of the face contour pixel data being calculated and face contour pixel data corresponding in reference picture, the similarity of face in detection picture and reference picture is determined.The invention also discloses a kind of detection devices of human face similarity degree.Using scheme disclosed by the invention, the convenience that human face similarity degree compares is improved, and improves the efficiency that human face similarity degree compares.
Description
Technical field
The present invention relates to ntelligent television technolog field more particularly to the detection methods and device of human face similarity degree.
Background technique
Due to the arrival in current intelligent epoch, the function of smart television is more and more intelligent, and camera has begun conduct
The standard configurations of TV, also, the idea of universal entertainment is also unconsciously being rooted in the hearts of the people, and everybody is concern star's
Relevant information, for example the news of the star liked is paid close attention to, or pay close attention to some position of the appearance of oneself and some part of star
It is whether similar.
Currently, user compares the face of oneself and the face of current star are more troublesome, for example, user needs to clap first
Under currently playing program star picture, compared with the face of the picture of oneself are then done naked eyes with the face of star,
In this way, the similarity of the human face five-sense-organ due to the face and star that are unable to judge accurately user, to cause user to face phase
Like degree comparison it is not convenient, and due to needing manually to go to compare, thus the time of user can be expended.
Above content is only used to facilitate the understanding of the technical scheme, and is not represented and is recognized that above content is existing skill
Art.
Summary of the invention
The convenience compared it is a primary object of the present invention to improve human face similarity degree, and improve human face similarity degree and compare
Efficiency.
To achieve the above object, the present invention provides a kind of detection method of human face similarity degree, comprising the following steps:
The face information in detection picture is obtained, the face information is stored as face contour pixel corresponding to face
Data;
The face contour pixel data are read according to preset read direction, are calculated by color sub-pixels described in reading
The difference value of face contour pixel data and face contour pixel data corresponding in reference picture;
According to the face contour pixel data being calculated and face contour pixel data corresponding in reference picture
Difference value, determine it is described detection picture and reference picture in face similarity.
Preferably, the face information obtained in detection picture, the face information is stored as corresponding to face
The step of face contour pixel data includes:
After system starting, the detection picture of active user is obtained, identifies and whether there is face information in the detection picture;
There are when face information in the detection picture, the face information is extracted, the face information is pressed into face
It is divided, and stores face contour pixel data corresponding to face respectively.
Preferably, the face contour pixel data are read according to preset read direction described, by color sub-pixels
Calculate the face contour pixel data and the step of the difference value of face contour pixel data corresponding in reference picture read
Before rapid, the detection method of the human face similarity degree further include:
Preset read direction is determined on the detection picture, reads all pixels on identified read direction
Point determines the face contour area in the detection picture, and obtains the face wire-frame image prime number in the face contour area
According to.
Preferably, the institute for determining preset read direction on detection picture, reading on identified read direction
There is pixel, determines the face contour area in the detection picture, and obtain the face profile in the face contour area
The step of pixel data includes:
Preset read direction is determined on detection picture, reads all pictures in the detection picture on the read direction
Vegetarian refreshments calculates the difference of the latter pixel and previous pixel;
When the difference of the latter pixel and previous pixel is greater than preset difference threshold for the first time, after determining this
One pixel is the starting point of face contour area, when the latter pixel and the difference of previous pixel are greater than for the second time
When preset difference threshold, determine that the latter pixel is the end point of face contour area, the starting point and end point
Between part be face contour area;
Obtain the corresponding face contour pixel data of all pixels point in the face contour area.
Preferably, described that the face contour pixel data are read according to preset read direction, based on color sub-pixels
Calculate the step of face contour pixel data read are with the difference value of face contour pixel data corresponding in reference picture
Include:
The face contour pixel data that the face contour area of the detection picture is read according to preset read direction, are pressed
Preset read direction reads the face contour pixel data of corresponding position in the reference picture;
Calculate separately the face of red pixel, green pixel and blue pixel in the face contour area of the detection picture
The difference value of contour pixel data and face contour pixel data corresponding in reference picture.
In addition, to achieve the above object, the present invention also provides a kind of detection device of human face similarity degree, the detection picture with
The detection device of reference picture human face similarity degree includes:
Module is obtained, for obtaining the face information in detection picture, the face information is stored as corresponding to face
Face contour pixel data;
Computing module, for reading the face contour pixel data according to preset read direction, by color sub-pixels
Calculate the face contour pixel data and the difference value of face contour pixel data corresponding in reference picture read;
Determining module, for according to the face contour pixel data being calculated and face corresponding in reference picture
The difference value of contour pixel data determines the similarity of face in the detection picture and reference picture.
Preferably, the acquisition module is specifically used for:
After system starting, the detection picture of active user is obtained, identifies and whether there is face information in the detection picture;
There are when face information in the detection picture, the face information is extracted, the face information is pressed into face
It is divided, and stores face contour pixel data corresponding to face respectively.
Preferably, the detection device of the detection picture and reference picture human face similarity degree further include:
Read module reads identified read direction for determining preset read direction on the detection picture
On all pixels point, determine it is described detection picture in face contour area, and obtain in the face contour area five
Official's contour pixel data.
Preferably, the read module is specifically used for:
Preset read direction is determined on detection picture, reads all pictures in the detection picture on the read direction
Vegetarian refreshments calculates the difference of the latter pixel and previous pixel;
When the difference of the latter pixel and previous pixel is greater than preset difference threshold for the first time, after determining this
One pixel is the starting point of face contour area, when the latter pixel and the difference of previous pixel are greater than for the second time
When preset difference threshold, determine that the latter pixel is the end point of face contour area, the starting point and end point
Between part be face contour area;
Obtain the corresponding face contour pixel data of all pixels point in the face contour area.
Preferably, the computing module is specifically used for:
The face contour pixel data of the face contour area of the detection picture are read according to preset read direction,
The face contour pixel data of corresponding position in the reference picture are read by preset read direction;
Calculate separately the face of red pixel, green pixel and blue pixel in the face contour area of the detection picture
The difference value of contour pixel data and face contour pixel data corresponding in reference picture.
The present invention extracts the face information in detection picture, by face information by the detection picture of acquisition active user
Face contour pixel data corresponding to face are stored as, face contour pixel data are read according to preset read direction, are pressed
Color sub-pixels calculate face contour pixel data and the difference of face contour pixel data corresponding in reference picture read
Value;According to the difference value of the face contour pixel data being calculated and face contour pixel data corresponding in reference picture,
Determine the similarity of face in detection picture and reference picture.By the face contour pixel data of reading detection picture, and with
Corresponding face contour pixel data are compared in reference picture, to obtain accurate similarity, without user's naked eyes
It goes to compare, to improve the convenience that human face similarity degree compares, and improves the efficiency that human face similarity degree compares.
Detailed description of the invention
Fig. 1 is the flow diagram of the detection method first embodiment of the present inventor's face similarity degree;
Fig. 2 is that will test the step that face information in picture is stored as face contour pixel data corresponding to face in Fig. 1
Rapid refinement flow diagram;
Fig. 3 is the flow diagram of the detection method second embodiment of the present inventor's face similarity degree;
Fig. 4 is to determine the face contour area in detection picture in Fig. 3 and obtain face wire-frame image in face contour area
Prime number according to the step of refinement flow diagram;
Fig. 5 is to calculate the face contour pixel data corresponding with reference picture five read by color sub-pixels in Fig. 1
The refinement flow diagram of the step of difference value of official's contour pixel data;
Fig. 6 is the functional block diagram of the detection device first embodiment of the present inventor's face similarity degree;
Fig. 7 is the functional block diagram of the detection device second embodiment of the present inventor's face similarity degree.
The embodiments will be further described with reference to the accompanying drawings for the realization, the function and the advantages of the object of the present invention.
Specific embodiment
It should be appreciated that the specific embodiments described herein are merely illustrative of the present invention, it is not used to limit this hair
It is bright.
The present invention provides a kind of detection method of human face similarity degree.
Referring to Fig.1, Fig. 1 is the flow diagram of the detection method first embodiment of the present inventor's face similarity degree.
In one embodiment, the detection method of the human face similarity degree includes:
Step S10 obtains the face information in detection picture, face information is stored as face profile corresponding to face
Pixel data;
The detection method of human face similarity degree provided by the present embodiment, can apply smart television or other can obtain
User detects picture and can be to the smart machine that picture is handled;Also, detection picture and ginseng involved in this method
The face in picture is examined by taking the face of yellow as an example.After the starting of smart machine system, external or smart machine can be passed through
Included photographic device obtains the detection picture of active user, and presses in the detection picture has been determined there are when face information
Video pictures are divided into region picture different corresponding to face by the face in face information, are such as divided into video pictures
Ear region picture, nasal area picture, eye areas picture, mouth region picture and brow region picture;Then by face
Corresponding each region picture is stored as corresponding face contour pixel data, and is respectively stored in five memories, this five
Official's contour pixel data include the rgb value of face part in each region picture, i.e. red pixel, green pixel and blue pixel
Tristimulus values, it is of course also possible to sort out to the tristimulus values stored and correspondence is deposited according to the profile and shape of face
Storage.
Step S20, reads face contour pixel data according to preset read direction, calculates reading by color sub-pixels
The difference value of face contour pixel data and face contour pixel data corresponding in reference picture;
When the face of active user in judgement detection picture are to the similarity of the corresponding face in reference picture, according to
Preset read direction reads the face contour pixel data in detection picture, which can be according to practical need
Customized setting is wanted, such as can be the starting point read with the half of the coordinate of current detection picture in the vertical direction, and
One-row pixels data are read in the horizontal direction, to determine the face contour area on the read direction, and in face profile region
Pixel data in domain is face contour pixel data.When relatively, the face profile of different color sub-pixels is only read
Pixel data, and by face contour pixel data corresponding to reference picture in the face contour pixel data of reading and database
It compares.Specially calculate the face contour pixel data corresponding with reference picture five of the different colours sub-pixel of reading
The difference value of official's contour pixel data;It can be to the corresponding face contour pixel data of each face stored respectively and with reference to figure
Piece carries out corresponding comparison, obtains face in the face contour pixel data and reference picture of the corresponding different pixels of each face
The difference value of contour pixel data.
In the present embodiment, reference picture can be the photograph of certain star character, and the storage form of the reference picture can be with
For face contour pixel data corresponding to star's face are respectively stored in corresponding memory, i.e., deposited at other five
Face contour pixel data corresponding to ear, nose, eyes, eyebrow and the mouth of the star are stored in reservoir respectively.It calculates and reads
The difference value of face contour pixel data and face contour pixel data corresponding in reference picture in the detection picture taken, such as
Calculate the red sub-pixel, green sub-pixels of ear profile or the pixel data of blue subpixels in detection picture and reference picture
Difference value, naturally it is also possible to be detect picture in user ear profile on all pixels point Hongyuan color value, green primary color value
Or the sum of blue primary value, subtract the Hongyuan color value, green primary color value of all pixels point on star character ear profile in reference picture
Or the sum of blue primary value, and the absolute value of obtained difference.
Step S30, according to the face contour pixel data being calculated and face contour pixel corresponding in reference picture
The difference value of data determines the similarity of face in detection picture and reference picture.
Obtain the face contour pixel data corresponding with reference picture five in the detection picture of different colours sub-pixel
After the difference value of official's contour pixel data, according to difference value, further determine that detection picture in reference picture face it is similar
Degree.In the present embodiment, obtained difference value can be summed, the human face five-sense-organ of reaction detection picture is carried out with obtained value
With the similarity of human face five-sense-organ in reference picture, the value is smaller, shows that the similarity of human face five-sense-organ is higher, it is, of course, also possible into
One step compares the sum acquired and preset threshold value, to judge similarity according to different standards.Detection picture is obtained
In middle human face five-sense-organ and reference picture after the similarity of human face five-sense-organ, face in active user and reference picture can be judged respectively
The similarity in each face region, certainly can also comprehensive descision active user it is whether similar to face in reference picture.
Active user may be used also with after the similarity in each face region of face in reference picture in having obtained detection picture
The size of the similarity is shown on the display screen of smart machine.Also, system can recommend quotient to user according to similarity
Product.
The present embodiment is extracted the face information in detection picture, face is believed by the detection picture of acquisition active user
Breath is stored as face contour pixel data corresponding to face, reads face contour pixel data according to preset read direction,
Face contour pixel data and the difference of face contour pixel data corresponding in reference picture read is calculated by color sub-pixels
Different value;According to the difference of the face contour pixel data being calculated and face contour pixel data corresponding in reference picture
Value determines the similarity of face in detection picture and reference picture.By reading the face contour pixel data of detection picture, and
Face contour pixel data corresponding with reference picture are compared, and to obtain accurate similarity, are not necessarily to user's meat
Eye goes to compare, to improve the convenience that human face similarity degree compares, and improves the efficiency that human face similarity degree compares.
The present embodiment can be used in purchase system or other service systems, once for example, user's face and certain star
When personage's face similarity is higher (such as ear similarity is higher), shopping or recommendation service can be recommended, it specifically can should
Star character represent or used the commodity with correlation or service be shown on a display screen, to recommend user,
Realize more preferably user experience.
Referring to Fig. 2, Fig. 2 is that will test face information in picture in Fig. 1 to be stored as face contour pixel corresponding to face
The refinement flow diagram of the step of data.
In the above-described embodiments, step S10 is specifically included:
Step S11 after system starting, obtains the detection picture of active user, whether there is face in recognition detection picture
Information;
Step S12 extracts face information there are when face information in detection picture, and face information is pressed face and is carried out
It divides, and stores face contour pixel data corresponding to face respectively.
After the starting of smart machine system, obtain active user's by the included photographic device of external or smart machine
Photo or video and as detection picture, the detection picture can be for currently using the one or more of viewing smart machine is current
Acquired in family is shot, and is identified by the method for image recognition and whether there is face information in obtained detection picture,
Judge detect picture in whether include people head and torso portion, equally, detect picture in may include one or more people
Face information.When in detection picture there are when face information, extracting face information, and according to the face in face information to detection
Picture is divided, and region picture corresponding to each face is divided into, then, by the corresponding face of each region picture
Contour pixel data are respectively stored in five different memories by color sub-pixels.
After system starting, the detection picture of active user is obtained, whether there is face information in recognition detection picture, work as inspection
There are when face information in mapping piece, face information is extracted, the face that will test picture to be divided into face information are corresponding
Region picture, and picture corresponding face contour pixel data in each region are stored by color sub-pixels, to improve people
The convenience that face similarity degree compares, and improve the efficiency that human face similarity degree compares and provide the foundation.
It is the flow diagram of the detection method second embodiment of the present inventor's face similarity degree referring to Fig. 3, Fig. 3.
Detection method first embodiment based on the present inventor's face similarity degree, before executing step S20, this method is also wrapped
It includes:
Step S40 determines preset read direction on detection picture, reads all pictures on identified read direction
Vegetarian refreshments determines the face contour area in detection picture, and obtains the face contour pixel data in face contour area.
In contrasting detection picture and reference picture before face similarity, pass through face chamfered shape symmetrical pixels first
Recognizer determines the face contour area in detection picture.Specifically, can detecte picture in the vertical direction coordinate two
Certain point at/mono- coordinate position is the starting point of read direction, the different colours sub-pixel of read level direction lastrow
Pixel data, such as from the central region of ear (by taking ear as an example) according to horizontal direction from left to right, successively calculate latter
(not needing will be equal between every two pixel of full line in the horizontal direction for the difference of the pixel data of pixel and previous pixel
It is compared, it is only necessary to be sampled as needed, the pixel that sampling obtains is compared), when former and later two pixels
When the difference of the pixel data of point is more than or equal to preset difference threshold, then the latter pixel is face contour area
Starting point;The difference for successively calculating the pixel data of latter pixel and previous pixel again, when occurring former and later two pictures again
When the difference of the pixel data of vegetarian refreshments is more than or equal to preset threshold values, then the previous pixel is the knot of face contour area
Beam spot.When determining preset read direction, the different colours sub-pixel for reading a line can be removed only according to a coordinate direction
Pixel data, the pixel data of the different colours sub-pixel of multirow on multiple coordinate directions also can be read.In starting point and knot
Region part between beam spot is the face contour area for detecting picture, it is determined that after the face contour area, wherein being wrapped
The pixel data of all pixels point contained is face contour pixel data.
With further reference to Fig. 4, Fig. 4 is to determine the face contour area in detection picture in Fig. 3 and obtain face profile region
Refinement flow diagram in domain the step of face contour pixel data.
In the present embodiment, step S40 is specifically included:
Step S41 determines preset read direction on detection picture, reads the institute in detection picture on the read direction
There is pixel, calculates the difference of the latter pixel and previous pixel;
Step S42, when the difference of the latter pixel and previous pixel is greater than preset difference threshold for the first time,
Determine that the latter pixel is the starting point of face contour area, when the difference the of the latter pixel and previous pixel
It is secondary when being greater than preset difference threshold, determine that the latter pixel is the end point of face contour area, starting point and knot
Part between beam spot is face contour area;
Step S43 obtains the corresponding face contour pixel data of all pixels point in face contour area.
When determining the face contour area in detection picture, preset read direction is determined on detection picture, such as with inspection
Mapping piece in the vertical direction the certain point at the half coordinate position of coordinate be read direction starting point, read level
The pixel data of the different colours sub-pixel of direction lastrow, and the pixel data of the pixel of reading is stored in memory
In.For example, it is assumed that the pixel data of first pixel in the readout direction are as follows: R1=0, G1=0, B1=0, by the point
Value is stored into second memory: MR1=0, MG1=0, MB1=0;The picture of the 20th pixel read in the readout direction
Prime number evidence are as follows: R20=0, G20=0, B20=0 store the value of the point into second memory: MR20=255, MG20=
255, MB20=255;The pixel data of the 40th pixel read in the readout direction are as follows: R40=0, G40=0, B40=
0, the value of the point is stored into second memory: MR40=255, MG40=255, MB40=255;It reads in the readout direction
The 60th pixel pixel data are as follows: R60=0, G60=0, B60=0 store the value of the point into second memory:
MR60=0, MG60=0, MB60=0.
The difference for calculating the pixel data of read the latter pixel and the pixel data of previous pixel, that is, use
The pixel data value of the latter pixel subtracts the pixel data value of previous pixel, obtained different colours sub-pixel
Difference is is respectively R_Dis, G_Dis and B_Dis, when the value of R_Dis, G_Dis and B_Dis are greater than preset difference threshold for the first time
When value, (by taking horizontal direction as an example) then can determine whether that the latter pixel is the starting point of face contour area, at this point, will starting
The value of point is stored into third memory, when the value of R_Dis, G_Dis and B_Dis are greater than preset difference threshold for the second time,
It then can determine whether that previous pixel is that the end point of face contour area equally stores the value of end point to third memory
In.In the present embodiment, preset difference threshold may be configured as 10.R_Dis, G_Dis and B_Dis are counted in the following manner respectively
It calculates:
According to starting point and end point, the face contour area in current detection picture can be determined, in the face profile
All pixels point in region is the effective pixel data of face contour area, obtains all pixels in face contour area
The corresponding face contour pixel data of point, and store into memory.
In contrasting detection picture and reference picture before face similarity, preset reading side is determined on detection picture
To reading detects all pixels point in picture on the read direction, calculates the difference of the latter pixel and previous pixel
Value;The starting point and end point for determining face contour area, the part between starting point and end point are face contour area, and
The corresponding face contour pixel data of all pixels point in face contour area are obtained, are further raising human face similarity degree ratio
Compared with convenience, and improve the efficiency that compares of human face similarity degree and provide the foundation.
It is to be calculated in the face contour pixel data and reference picture read in Fig. 1 by color sub-pixels referring to Fig. 5, Fig. 5
The refinement flow diagram of the step of difference value of corresponding face contour pixel data.
The first, second embodiment of detection method based on aforementioned present invention human face similarity degree, step S20 are specifically included:
Step S21 reads the face wire-frame image prime number of the face contour area of detection picture according to preset read direction
According to by the face contour pixel data of corresponding position in preset read direction reading reference picture;
Step S22 calculates separately red pixel in the face contour area of detection picture, green pixel and blue pixel
The difference value of face contour pixel data and face contour pixel data corresponding in reference picture.
After the order that system gets red sub-pixel similarity-rough set, according to preset read direction, detection is read
The face contour pixel data of red pixel in the face contour area of picture, using the face contour pixel data as input number
According to, while the face contour pixel data of corresponding position in reference picture are read as reference data.Below to detect in picture
For region picture corresponding to ear, the calculation method of the difference value of face contour pixel data is illustrated:
Assuming that region picture corresponding to ear is by E1R1, E1G1, the color color matching of E1B1 pixel is into (E1 indicates ear
Region picture corresponding to piece, R1, G1, B1 indicate pixel RGB, i.e. RGB), in reference picture corresponding ear picture be by
The color of E2R1, E2G1, E2B1 pixel is matched colors.Due to yellow skin red pixel range in R=221-255
It is interior, the face contour pixel of reads pixel data value in the range in the face contour pixel data of the detection picture of storage
Data and sum obtain, while the pixel data value of corresponding position in reference picture is read in the range
Interior face contour pixel data and summation obtain;Then, all red sons of detection picture are calculated
The value of face contour pixel data summation in pixel coverage, with the face wheel within the scope of red sub-pixels all in reference picture
The difference of the value of wide pixel data summation, i.e.,It obtains in detection picture and reference picture
The difference value of the face contour pixel data of red sub-pixel.Wherein, the size of n goes to determine according to the pixel of entire picture,
The present embodiment by taking n=10 as an example, thus it can be assumed that:
And E_R_total=| 2210-2200 |=10.That is, detecting the face wire-frame image of the red sub-pixel of ear in picture
Prime number is 10 according to the difference value of the face contour pixel data of the red sub-pixel of ear corresponding in reference picture.This implementation
In example, when the range of difference value is within 0-1000, the face in the detection picture of system record input and phase in reference picture
The face answered are similar;Difference value is smaller, such as E_R_total=0, illustrates that the face detected in picture are corresponding to reference picture
Face similarity it is higher, difference value is bigger, such as E_R_total=1000, illustrate detect picture in face and reference
The similarity of corresponding face is lower in picture.
After the order that system gets green sub-pixels similarity-rough set, according to preset read direction, read each
The face contour pixel data for detecting the face contour area Green pixel of picture, using the face contour pixel data as defeated
Enter data, while reading the face contour pixel data of corresponding position in reference picture as reference data.Below equally with inspection
In mapping piece for region picture corresponding to ear, the calculation method of the difference value of face contour pixel data is said
It is bright:
Assuming that region picture corresponding to ear is by E1R1, E1G1, the color color matching of E1B1 pixel is into (E1 indicates ear
Region picture corresponding to piece, R1, G1, B1 indicate pixel RGB, i.e. RGB), in reference picture corresponding ear picture be by
The color of E2R1, E2G1, E2B1 pixel is matched colors.Due to yellow skin red sub-pixel range in G=154-176
It is interior, the face contour pixel of reads pixel data value in the range in the face contour pixel data of the detection picture of storage
Data and sum obtainThe pixel data value of corresponding position in reference picture is read simultaneously in the range
Face contour pixel data and sum obtainThen, all green sub-pixels models of detection picture are calculated
The value of face contour pixel data summation in enclosing, with the face contour pixel within the scope of green sub-pixels all in reference picture
The difference of the value of data summation, i.e.,Obtain detection picture and reference picture Green
The difference value of the face contour pixel data of sub-pixel.Wherein, the size of n goes to determine according to the pixel of entire picture, this reality
Example is applied by taking n=10 as an example, thus it can be assumed that:And E_
G_total=| 1650-1600 |=50.That is, detecting the face wire-frame image prime number of the green sub-pixels of ear in picture
It is 50 according to the difference value of the face contour pixel data of the green sub-pixels of face corresponding in reference picture.The present embodiment
In, when the range of difference value is within 0-1000, the face in the detection picture of system record input are corresponding to reference picture
Face it is similar;Difference value is smaller, such as E_G_total=0, illustrates that the face detected in picture are corresponding with reference picture
The similarity of face is higher, and difference value is bigger, such as E_G_total=1000, illustrate detect picture in face with reference to figure
The similarity of corresponding face is lower in piece.
After the order that system gets blue subpixels similarity-rough set, according to preset read direction, read each
The face contour pixel data for detecting blue pixel in the face contour area of picture, using the face contour pixel data as defeated
Enter data, while reading the face contour pixel data of corresponding position in reference picture as reference data.Below equally with inspection
In mapping piece for region picture corresponding to ear, the calculation method of the difference value of face contour pixel data is said
It is bright:
Assuming that region picture corresponding to ear is by E1R1, E1G1, the color color matching of E1B1 pixel is into (E1 indicates ear
Region picture corresponding to piece, R1, G1, B1 indicate pixel RGB, i.e. RGB), in reference picture corresponding ear picture be by
The color of E2R1, E2G1, E2B1 pixel is matched colors.Due to yellow skin blue subpixels range in G=127-145
It is interior, the face contour pixel of reads pixel data value in the range in the face contour pixel data of the detection picture of storage
Data and sum obtainRead the pixel data value of corresponding position in reference picture in the range simultaneously
Face contour pixel data and sum obtainThen, all blue subpixels models of detection picture are calculated
The value of face contour pixel data summation in enclosing, with the face contour pixel within the scope of blue subpixels all in reference picture
The difference of the value of data summation, i.e.,Obtain blue in detection picture and reference picture
The difference value of the face contour pixel data of sub-pixel.Wherein, the size of n goes to determine according to the pixel of entire picture, this reality
Example is applied by taking n=10 as an example, thus it can be assumed that:And
E_B_total=| 1400-1300 |=100.That is, detecting the face contour pixel of the blue subpixels of ear in picture
Data are 100 with the difference value of the face contour pixel data of the blue subpixels of face corresponding in reference picture.This implementation
In example, when the range of difference value is within 0-1000, the face in the detection picture of system record input and phase in reference picture
The face answered are similar;Difference value is smaller, such as E_B_total=0, illustrates that the face detected in picture are corresponding to reference picture
Face similarity it is higher, difference value is bigger, such as E_B_total=1000, illustrate detect picture in face and reference
The similarity of corresponding face is lower in picture.
Red sub-pixel, green sub-pixels and blue subpixels in the face contour area for having respectively obtained detection picture
Face contour pixel data with after the difference value of face contour pixel data corresponding in reference picture, to these three difference values
The similarity E_C_total=of face in detection picture and reference picture can be obtained in summation | E_R_total+E_G_total+
E_B_total |=| 10+50+100 |=160.That is, three kinds of the face part of the yellow in current detection picture face
Sub-pixels and the similarity of three kinds of color sub-pixels face of picture yellow of reference are E_C_total=160, equally, should
It is worth smaller, shows that similarity is higher.
The face contour pixel data in the face contour area of detection picture are read according to preset read direction, by pre-
If read direction read reference picture in corresponding position face contour pixel data;Calculate separately the face wheel of detection picture
The face contour pixel data of red sub-pixel, green sub-pixels and blue subpixels are corresponding with reference picture in wide region
The difference value of face contour pixel data further ensures and can be improved the convenience that human face similarity degree compares, and improves
The efficiency that human face similarity degree compares.
The present invention also provides a kind of detection devices of human face similarity degree.
It is the functional block diagram of the detection device first embodiment of the present inventor's face similarity degree referring to Fig. 6, Fig. 6.
In one embodiment, the detection device of human face similarity degree includes:
Module 10 is obtained, for obtaining the face information in detection picture, face information is stored as corresponding to face
Face contour pixel data;
Computing module 20, for reading face contour pixel data according to preset read direction, based on color sub-pixels
Calculate face contour pixel data and the difference value of face contour pixel data corresponding in reference picture read;
Determining module 30, for according to the face contour pixel data being calculated and face wheel corresponding in reference picture
The difference value of wide pixel data determines the similarity of face in detection picture and reference picture.
The detection device of human face similarity degree provided by the present embodiment, can apply smart television or other can obtain
User detects picture and can be to the smart machine that picture is handled;Also, in related detection picture and reference picture
Face by taking the face of yellow as an example.After the starting of smart machine system, obtaining module 10 can be set by external or intelligence
Standby included photographic device obtains the detection picture of active user, and there are when face information in the detection picture has been determined,
Video pictures are divided into region picture different corresponding to face by the face in face information, such as divide video pictures
For ear region picture, nasal area picture, eye areas picture, mouth region picture and brow region picture;Then by five
Each region picture corresponding to official is stored as corresponding face contour pixel data, and is respectively stored in five memories, should
Face contour pixel data include the rgb value of face part in each region picture, i.e. red pixel, green pixel and blue pixel
Tristimulus values the tristimulus values stored sort out and correspondence is deposited it is of course also possible to according to the profiles and shape of face
Storage.
When the face of active user in judgement detection picture are to the similarity of the corresponding face in reference picture, calculate
Module 20 reads the face contour pixel data in detection picture, which can according to preset read direction
Customized setting according to actual needs, as that can be with the half of the coordinate of current detection picture in the vertical direction to read
Starting point, and one-row pixels data are read in the horizontal direction, to determine the face contour area on the read direction, and five
Pixel data in official's contour area is face contour pixel data.When relatively, different color sub-pixels are only read
Face contour pixel data, and by face profile corresponding to reference picture in the face contour pixel data of reading and database
Pixel data compares.It specially calculates in the face contour pixel data and reference picture of the different colours sub-pixel of reading
The difference value of corresponding face contour pixel data;The corresponding face contour pixel data of each face stored can be distinguished
Compared with reference picture correspond to, the face contour pixel data and reference picture of the corresponding different pixels of each face are obtained
The difference value of middle face contour pixel data.
In the present embodiment, reference picture can be the photograph of certain star character, and the storage form of the reference picture can be with
For face contour pixel data corresponding to star's face are respectively stored in corresponding memory, i.e., deposited at other five
Face contour pixel data corresponding to ear, nose, eyes, eyebrow and the mouth of the star are stored in reservoir respectively.It calculates and reads
The difference value of face contour pixel data and face contour pixel data corresponding in reference picture in the detection picture taken, such as
Calculate the red sub-pixel, green sub-pixels of ear profile or the pixel data of blue subpixels in detection picture and reference picture
Difference value, naturally it is also possible to be detect picture in user ear profile on all pixels point Hongyuan color value, green primary color value
Or the sum of blue primary value, subtract the Hongyuan color value, green primary color value of all pixels point on star character ear profile in reference picture
Or the sum of blue primary value, and the absolute value of obtained difference.
Obtain the face contour pixel data corresponding with reference picture five in the detection picture of different colours sub-pixel
After the difference value of official's contour pixel data, determining module 30 further determines that in detection picture and reference picture according to difference value
The similarity of face.In the present embodiment, obtained difference value can be summed, reaction detection picture is come with obtained value
Human face five-sense-organ and reference picture in human face five-sense-organ similarity, the value is smaller, show that the similarity of human face five-sense-organ is higher, when
So, further the sum acquired and preset threshold value can also be compared, to judge similarity according to different standards.It obtains
Active user and reference can be judged respectively after the similarity of human face five-sense-organ in human face five-sense-organ and reference picture in detection picture
The similarity in each face region of face in picture, certainly can also comprehensive descision active user and face in reference picture whether phase
Seemingly.
Active user may be used also with after the similarity in each face region of face in reference picture in having obtained detection picture
The size of the similarity is shown on the display screen of smart machine.Also, system can recommend quotient to user according to similarity
Product.
The present embodiment extracts the face information in video pictures, by video figure by the video pictures of acquisition active user
Piece is stored as corresponding detection picture by the face in face information, and the face of detection picture are read according to preset read direction
Contour pixel data are calculated face contour pixel data and the face profile corresponding in reference picture read by color sub-pixels
The difference value of pixel data;It is corresponding with reference picture to the face contour pixel data of the corresponding reading of different colours sub-pixel
Face contour pixel data difference value summation, obtain detection picture and reference picture in face similarity.Pass through reading
The face contour pixel data of picture are detected, and are compared with face contour pixel data corresponding in reference picture, to obtain
Accurate similarity is obtained, goes to compare without user's naked eyes, to improve the convenience that human face similarity degree compares, and improve
The efficiency that human face similarity degree compares.
The present embodiment can be used in purchase system or other service systems, once for example, user's face and certain star
When personage's face similarity is higher (such as ear similarity is higher), shopping or recommendation service can be recommended, it specifically can should
Star character represent or used the commodity with correlation or service be shown on a display screen, to recommend user,
Realize more preferably user experience.
In above-mentioned implementation column, obtains module 10 and is specifically used for:
After system starting, the detection picture of active user is obtained, whether there is face information in recognition detection picture;
There are when face information in detection picture, face information is extracted, face information is divided by face, and point
It Cun Chu not face contour pixel data corresponding to face.
After the starting of smart machine system, obtains module 10 and obtained by the included photographic device of external or smart machine
The photo or video of active user and as detection picture, the detection picture can for currently in viewing one of smart machine or
Acquired in multiple active users are shot, and is identified in obtained detection picture and be whether there is by the method for image recognition
Face information, that is, judge detect picture in whether include people head and torso portion, equally, detect picture in may include one
Or multiple face informations.When in detection picture there are when face information, extracting face information, and according to five in face information
Official divides detection picture, region picture corresponding to each face is divided into, then, by each region picture
Corresponding face contour pixel data are respectively stored in five different memories by color sub-pixels.
After system starting, the detection picture of active user is obtained, whether there is face information in recognition detection picture, work as inspection
There are when face information in mapping piece, face information is extracted, the face that will test picture to be divided into face information are corresponding
Region picture, and picture corresponding face contour pixel data in each region are stored by color sub-pixels, to improve people
The convenience that face similarity degree compares, and improve the efficiency that human face similarity degree compares and provide the foundation.
It is the functional block diagram of the detection device second embodiment of the present inventor's face similarity degree referring to Fig. 7, Fig. 7.
On the basis of the detection device first embodiment of aforementioned present invention human face similarity degree, the device further include:
Read module 40 is read on identified read direction for determining preset read direction on detection picture
All pixels point, determine the face contour area in detection picture, and obtain the face contour pixel in face contour area
Data.
In contrasting detection picture and reference picture before face similarity, pass through face chamfered shape symmetrical pixels first
Recognizer determines the face contour area in detection picture.Specifically, read module 40 can detecte picture in vertical direction
Certain point at the half coordinate position of upper coordinate is the starting point of read direction, the difference of read level direction lastrow
The pixel data of color sub-pixels, for example, from the central region of ear (by taking ear as an example) according to horizontal direction from left to right, according to
The difference of the secondary pixel data for calculating latter pixel and previous pixel (is not needed every two picture of full line in the horizontal direction
It is compared between vegetarian refreshments, it is only necessary to be sampled as needed, the pixel that sampling obtains is compared), currently
When the difference of the pixel data of latter two pixel is more than or equal to preset difference threshold, then the latter pixel is face
The starting point of contour area;The difference for successively calculating the pixel data of latter pixel and previous pixel again, when occurring again
When the difference of the pixel data of former and later two pixels is more than or equal to preset threshold values, then the previous pixel is face wheel
The end point in wide region.When determining preset read direction, the difference for reading a line can be gone only according to a coordinate direction
The pixel data of the different colours sub-pixel of multirow on multiple coordinate directions also can be read in the pixel data of color sub-pixels.?
Region part between starting point and end point is the face contour area for detecting picture, it is determined that the face contour area
Afterwards, the pixel data of all pixels point included in is face contour pixel data.
In the present embodiment, read module 40 is specifically used for:
Preset read direction is determined on detection picture, reads all pixels in detection picture on the read direction
Point calculates the difference of the latter pixel and previous pixel;
When the difference of the latter pixel and previous pixel is greater than preset difference threshold for the first time, after determining this
One pixel is the starting point of face contour area, when the latter pixel and the difference of previous pixel are greater than for the second time
When preset difference threshold, determine that the latter pixel is the end point of face contour area, between starting point and end point
Part be face contour area;
Obtain the corresponding face contour pixel data of all pixels point in face contour area.
When determining the face contour area in detection picture, preset read direction is determined on detection picture, such as with inspection
Mapping piece in the vertical direction the certain point at the half coordinate position of coordinate be read direction starting point, read level
The pixel data of the different colours sub-pixel of direction lastrow, and the pixel data of the pixel of reading is stored in memory
In.For example, it is assumed that the pixel data of first pixel in the readout direction are as follows: R1=0, G1=0, B1=0, by the point
Value is stored into second memory: MR1=0, MG1=0, MB1=0;The picture of the 20th pixel read in the readout direction
Prime number evidence are as follows: R20=0, G20=0, B20=0 store the value of the point into second memory: MR20=255, MG20=
255, MB20=255;The pixel data of the 40th pixel read in the readout direction are as follows: R40=0, G40=0, B40=
0, the value of the point is stored into second memory: MR40=255, MG40=255, MB40=255;It reads in the readout direction
The 60th pixel pixel data are as follows: R60=0, G60=0, B60=0 store the value of the point into second memory:
MR60=0, MG60=0, MB60=0.
The difference for calculating the pixel data of read the latter pixel and the pixel data of previous pixel, that is, use
The pixel data value of the latter pixel subtracts the pixel data value of previous pixel, obtained different colours sub-pixel
Difference is is respectively R_Dis, G_Dis and B_Dis, when the value of R_Dis, G_Dis and B_Dis are greater than preset difference threshold for the first time
When value, (by taking horizontal direction as an example) then can determine whether that the latter pixel is the starting point of face contour area, at this point, will starting
The value of point is stored into third memory, when the value of R_Dis, G_Dis and B_Dis are greater than preset difference threshold for the second time,
It then can determine whether that previous pixel is that the end point of face contour area equally stores the value of end point to third memory
In.In the present embodiment, preset difference threshold may be configured as 10.R_Dis, G_Dis and B_Dis are counted in the following manner respectively
It calculates:
According to starting point and end point, the face contour area in current detection picture can be determined, in the face profile
All pixels point in region is the effective pixel data of face contour area, obtains all pixels in face contour area
The corresponding face contour pixel data of point, and store into memory.
In contrasting detection picture and reference picture before face similarity, preset reading side is determined on detection picture
To reading detects all pixels point in picture on the read direction, calculates the difference of the latter pixel and previous pixel
Value;The starting point and end point for determining face contour area, the part between starting point and end point are face contour area, and
The corresponding face contour pixel data of all pixels point in face contour area are obtained, are further raising human face similarity degree ratio
Compared with convenience, and improve the efficiency that compares of human face similarity degree and provide the foundation.
The first, second embodiment of detection device based on aforementioned present invention human face similarity degree, computing module 20 are specifically used for:
The face contour pixel data that the face contour area of detection picture is read according to preset read direction, by default
Read direction read reference picture in corresponding position face contour pixel data;
Calculate separately the face profile of red pixel, green pixel and blue pixel in the face contour area of detection picture
The difference value of pixel data and face contour pixel data corresponding in reference picture.
After the order that system gets red sub-pixel similarity-rough set, according to preset read direction, detection is read
The face contour pixel data of red pixel in the face contour area of picture, using the face contour pixel data as input number
According to, while the face contour pixel data of corresponding position in reference picture are read as reference data.Below to detect in picture
For region picture corresponding to ear, the calculation method of the difference value of face contour pixel data is illustrated:
Assuming that region picture corresponding to ear is by E1R1, E1G1, the color color matching of E1B1 pixel is into (E1 indicates ear
Region picture corresponding to piece, R1, G1, B1 indicate pixel RGB, i.e. RGB), in reference picture corresponding ear picture be by
The color of E2R1, E2G1, E2B1 pixel is matched colors.Due to yellow skin red pixel range in R=221-255
It is interior, the face contour pixel of reads pixel data value in the range in the face contour pixel data of the detection picture of storage
Data and sum obtainRead the pixel data value of corresponding position in reference picture in the range simultaneously
Face contour pixel data and sum obtainThen, all red sub-pixel models of detection picture are calculated
The value of face contour pixel data summation in enclosing, with the face contour pixel within the scope of red sub-pixels all in reference picture
The difference of the value of data summation, i.e.,Obtain red sub- picture in detection picture and reference picture
The difference value of the face contour pixel data of element.Wherein, the size of n goes to determine according to the pixel of entire picture, the present embodiment
By taking n=10 as an example, thus it can be assumed that:And
E_R_total=| 2210-2200 |=10.That is, detecting the face contour pixel of the red sub-pixel of ear in picture
Data are 10 with the difference value of the face contour pixel data of the red sub-pixel of ear corresponding in reference picture.The present embodiment
In, when the range of difference value is within 0-1000, the face in the detection picture of system record input are corresponding to reference picture
Face it is similar;Difference value is smaller, such as E_R_total=0, illustrates that the face detected in picture are corresponding with reference picture
The similarity of face is higher, and difference value is bigger, such as E_R_total=1000, illustrate detect picture in face with reference to figure
The similarity of corresponding face is lower in piece.
After the order that system gets green sub-pixels similarity-rough set, according to preset read direction, read each
The face contour pixel data for detecting the face contour area Green pixel of picture, using the face contour pixel data as defeated
Enter data, while reading the face contour pixel data of corresponding position in reference picture as reference data.Below equally with inspection
In mapping piece for region picture corresponding to ear, the calculation method of the difference value of face contour pixel data is said
It is bright:
Assuming that region picture corresponding to ear is by E1R1, E1G1, the color color matching of E1B1 pixel is into (E1 indicates ear
Region picture corresponding to piece, R1, G1, B1 indicate pixel RGB, i.e. RGB), in reference picture corresponding ear picture be by
The color of E2R1, E2G1, E2B1 pixel is matched colors.Due to yellow skin red sub-pixel range in G=154-176
It is interior, the face contour pixel of reads pixel data value in the range in the face contour pixel data of the detection picture of storage
Data and sum obtainRead the pixel data value of corresponding position in reference picture in the range simultaneously
Face contour pixel data and sum obtainThen, all green sub-pixels of detection picture are calculated
The value of face contour pixel data summation in range, with the face wire-frame image within the scope of green sub-pixels all in reference picture
Prime number according to the value of summation difference, i.e.,Obtain detection picture and reference picture
The difference value of the face contour pixel data of Green sub-pixel.Wherein, the size of n goes to determine according to the pixel of entire picture
It is fixed, the present embodiment by taking n=10 as an example, thus it can be assumed that:And E_
G_total=| 1650-1600 |=50.That is, detecting the face wire-frame image prime number of the green sub-pixels of ear in picture
It is 50 according to the difference value of the face contour pixel data of the green sub-pixels of face corresponding in reference picture.The present embodiment
In, when the range of difference value is within 0-1000, the face in the detection picture of system record input are corresponding to reference picture
Face it is similar;Difference value is smaller, such as E_G_total=0, illustrates that the face detected in picture are corresponding with reference picture
The similarity of face is higher, and difference value is bigger, such as E_G_total=1000, illustrate detect picture in face with reference to figure
The similarity of corresponding face is lower in piece.
After the order that system gets blue subpixels similarity-rough set, according to preset read direction, read each
The face contour pixel data for detecting blue pixel in the face contour area of picture, using the face contour pixel data as defeated
Enter data, while reading the face contour pixel data of corresponding position in reference picture as reference data.Below equally with inspection
In mapping piece for region picture corresponding to ear, the calculation method of the difference value of face contour pixel data is said
It is bright:
Assuming that region picture corresponding to ear is by E1R1, E1G1, the color color matching of E1B1 pixel is into (E1 indicates ear
Region picture corresponding to piece, R1, G1, B1 indicate pixel RGB, i.e. RGB), in reference picture corresponding ear picture be by
The color of E2R1, E2G1, E2B1 pixel is matched colors.Due to yellow skin blue subpixels range in G=127-145
It is interior, the face contour pixel of reads pixel data value in the range in the face contour pixel data of the detection picture of storage
Data and sum obtainThe pixel data value of corresponding position in reference picture is read simultaneously in the range
Face contour pixel data and sum obtainThen, all blue subpixels ranges of detection picture are calculated
The value of interior face contour pixel data summation, with the face wire-frame image prime number within the scope of blue subpixels all in reference picture
According to the difference of the value of summation, i.e.,Obtain blue in detection picture and reference picture
The difference value of the face contour pixel data of sub-pixel.Wherein, the size of n goes to determine according to the pixel of entire picture, this reality
Example is applied by taking n=10 as an example, thus it can be assumed that:
And E_B_total=| 1400-1300 |=100.That is, detecting the face wire-frame image of the blue subpixels of ear in picture
Prime number is 100 according to the difference value of the face contour pixel data of the blue subpixels of face corresponding in reference picture.This reality
It applies in example, when the range of difference value is within 0-1000, in the face and reference picture in the detection picture of system record input
Corresponding face are similar;Difference value is smaller, such as E_B_total=0, illustrates to detect the face in picture and phase in reference picture
The similarity for the face answered is higher, and difference value is bigger, such as E_B_total=1000, illustrates to detect face and ginseng in picture
The similarity for examining corresponding face in picture is lower.
Red sub-pixel, green sub-pixels and the sub- picture of blue in the face contour area for having respectively obtained detection picture
Element face contour pixel data with after the difference value of face contour pixel data corresponding in reference picture, to these three differences
Value summation, can be obtained the similarity E_C_total=of face in detection picture and reference picture | E_R_total+E_G_
Total+E_B_total |=| 10+50+100 |=160.That is, the face part of the yellow in current detection picture
The similarity of three kinds of color sub-pixels and three kinds of color sub-pixels face of picture yellow of reference is E_C_total=160, together
Sample, the value is smaller, shows that similarity is higher.
The face contour pixel data in the face contour area of detection picture are read according to preset read direction, by pre-
If read direction read reference picture in corresponding position face contour pixel data;Calculate separately the face wheel of detection picture
The face contour pixel data of red sub-pixel, green sub-pixels and blue subpixels are corresponding with reference picture in wide region
The difference value of face contour pixel data further ensures and can be improved the convenience that human face similarity degree compares, and improves
The efficiency that human face similarity degree compares.
The above is only a preferred embodiment of the present invention, is not intended to limit the scope of the invention, all to utilize this hair
Equivalent structure or equivalent flow shift made by bright specification and accompanying drawing content is applied directly or indirectly in other relevant skills
Art field, is included within the scope of the present invention.
Claims (6)
1. a kind of detection method of human face similarity degree, which comprises the following steps:
The face information in detection picture is obtained, the face information is stored as face wire-frame image prime number corresponding to face
According to;
Preset read direction is determined on detection picture, reads all pixels in the detection picture on the read direction
Point calculates the difference of the latter pixel and previous pixel;
When the difference of the latter pixel and previous pixel is greater than preset difference threshold for the first time, the latter is determined
Pixel be face contour area starting point, when the difference of the latter pixel and previous pixel be greater than for the second time it is default
Difference threshold when, determine the latter pixel be face contour area end point, between the starting point and end point
Part be face contour area;
Obtain the corresponding face contour pixel data of all pixels point in the face contour area;
The face contour pixel data are read according to preset read direction, the face read are calculated by color sub-pixels
The difference value of contour pixel data and face contour pixel data corresponding in reference picture;
According to the difference of the face contour pixel data being calculated and face contour pixel data corresponding in reference picture
Different value determines the similarity of face in the detection picture and reference picture.
2. the detection method of human face similarity degree as described in claim 1, which is characterized in that the people obtained in detection picture
Face information, the step of face information is stored as face contour pixel data corresponding to face include:
After system starting, the detection picture of active user is obtained, identifies and whether there is face information in the detection picture;
There are when face information in the detection picture, the face information is extracted, the face information is pressed into face and is carried out
It divides, and stores face contour pixel data corresponding to face respectively.
3. the detection method of human face similarity degree as described in claim 1, which is characterized in that described according to preset read direction
The face contour pixel data are read, the face contour pixel data and reference picture read are calculated by color sub-pixels
In corresponding face contour pixel data difference value the step of include:
The face contour pixel data that the face contour area of the detection picture is read according to preset read direction, by default
Read direction read the face contour pixel data of corresponding position in the reference picture;
Calculate separately the face profile of red pixel, green pixel and blue pixel in the face contour area of the detection picture
The difference value of pixel data and face contour pixel data corresponding in reference picture.
4. a kind of detection device of human face similarity degree, which is characterized in that the detection device of the human face similarity degree includes:
Module is obtained, for obtaining the face information in detection picture, the face information is stored as five corresponding to face
Official's contour pixel data;
Read module is used for:
Preset read direction is determined on detection picture, reads all pixels in the detection picture on the read direction
Point calculates the difference of the latter pixel and previous pixel;
When the difference of the latter pixel and previous pixel is greater than preset difference threshold for the first time, the latter is determined
Pixel be face contour area starting point, when the difference of the latter pixel and previous pixel be greater than for the second time it is default
Difference threshold when, determine the latter pixel be face contour area end point, between the starting point and end point
Part be face contour area;
Obtain the corresponding face contour pixel data of all pixels point in the face contour area;
Computing module is calculated for reading the face contour pixel data according to preset read direction by color sub-pixels
The face contour pixel data and the difference value of face contour pixel data corresponding in reference picture read;
Determining module, for according to the face contour pixel data being calculated and face profile corresponding in reference picture
The difference value of pixel data determines the similarity of face in the detection picture and reference picture.
5. the detection device of human face similarity degree as claimed in claim 4, which is characterized in that the acquisition module is specifically used for:
After system starting, the detection picture of active user is obtained, identifies and whether there is face information in the detection picture;
There are when face information in the detection picture, the face information is extracted, the face information is pressed into face and is carried out
It divides, and stores face contour pixel data corresponding to face respectively.
6. the detection device of human face similarity degree as claimed in claim 4, which is characterized in that the computing module is specifically used for:
The face contour pixel data that the face contour area of the detection picture is read according to preset read direction, by default
Read direction read the face contour pixel data of corresponding position in the reference picture;
Calculate separately the face profile of red pixel, green pixel and blue pixel in the face contour area of the detection picture
The difference value of pixel data and face contour pixel data corresponding in reference picture.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201410306792.5A CN105303151B (en) | 2014-06-30 | 2014-06-30 | The detection method and device of human face similarity degree |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201410306792.5A CN105303151B (en) | 2014-06-30 | 2014-06-30 | The detection method and device of human face similarity degree |
Publications (2)
Publication Number | Publication Date |
---|---|
CN105303151A CN105303151A (en) | 2016-02-03 |
CN105303151B true CN105303151B (en) | 2019-02-26 |
Family
ID=55200402
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201410306792.5A Active CN105303151B (en) | 2014-06-30 | 2014-06-30 | The detection method and device of human face similarity degree |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN105303151B (en) |
Families Citing this family (10)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN106096520A (en) * | 2016-06-02 | 2016-11-09 | 乐视控股(北京)有限公司 | A kind of information-pushing method and device |
CN106127974B (en) * | 2016-06-27 | 2018-08-03 | 东软集团股份有限公司 | A kind of the something lost card based reminding method and device of self-help terminal equipment |
CN106815309A (en) * | 2016-12-20 | 2017-06-09 | 北京奇虎科技有限公司 | A kind of image method for pushing, device and mobile terminal |
CN107032223A (en) * | 2017-05-04 | 2017-08-11 | 苏州大学 | The self-start method and its self-starting device of a kind of staircase |
CN107480628B (en) * | 2017-08-10 | 2020-08-25 | 苏州大学 | Face recognition method and device |
CN110097470A (en) * | 2019-03-19 | 2019-08-06 | 深圳壹账通智能科技有限公司 | Finance product promotion method, device and computer equipment based on artificial intelligence |
CN112351327A (en) * | 2019-08-06 | 2021-02-09 | 北京字节跳动网络技术有限公司 | Face image processing method and device, terminal and storage medium |
CN111147891B (en) * | 2019-12-31 | 2022-09-13 | 杭州威佩网络科技有限公司 | Method, device and equipment for acquiring information of object in video picture |
CN111230862B (en) * | 2020-01-10 | 2021-05-04 | 上海发那科机器人有限公司 | Handheld workpiece deburring method and system based on visual recognition function |
CN116071804A (en) * | 2023-01-18 | 2023-05-05 | 北京六律科技有限责任公司 | Face recognition method and device and electronic equipment |
Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN203054719U (en) * | 2012-11-11 | 2013-07-10 | 西安天动数字科技有限公司 | Face recognition interactive entertainment system |
CN103207898A (en) * | 2013-03-19 | 2013-07-17 | 天格科技(杭州)有限公司 | Method for rapidly retrieving similar faces based on locality sensitive hashing |
CN103456010A (en) * | 2013-09-02 | 2013-12-18 | 电子科技大学 | Human face cartoon generation method based on feature point localization |
Family Cites Families (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US8542928B2 (en) * | 2005-09-26 | 2013-09-24 | Canon Kabushiki Kaisha | Information processing apparatus and control method therefor |
-
2014
- 2014-06-30 CN CN201410306792.5A patent/CN105303151B/en active Active
Patent Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN203054719U (en) * | 2012-11-11 | 2013-07-10 | 西安天动数字科技有限公司 | Face recognition interactive entertainment system |
CN103207898A (en) * | 2013-03-19 | 2013-07-17 | 天格科技(杭州)有限公司 | Method for rapidly retrieving similar faces based on locality sensitive hashing |
CN103456010A (en) * | 2013-09-02 | 2013-12-18 | 电子科技大学 | Human face cartoon generation method based on feature point localization |
Non-Patent Citations (1)
Title |
---|
大变形条件下的人脸识别研究;张龙媛;《中国优秀硕士学位论文全文数据库 信息科技辑》;20131215;第1-47页 |
Also Published As
Publication number | Publication date |
---|---|
CN105303151A (en) | 2016-02-03 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN105303151B (en) | The detection method and device of human face similarity degree | |
CN105338887B (en) | The sensory evaluation device of skin and the evaluation method of skin | |
CN109583285B (en) | Object recognition method | |
US20230298383A1 (en) | Spoofing detection device, spoofing detection method, and recording medium | |
WO2020207423A1 (en) | Skin type detection method, skin type grade classification method and skin type detection apparatus | |
US20170153711A1 (en) | Visual Language for Human Computer Interfaces | |
US20140354947A1 (en) | Virtual glasses try-on method and apparatus thereof | |
CN110251066A (en) | Based on the not positive system and method for subjective distance measuring measurement ophthalmic refractive | |
CN110032915A (en) | A kind of human face in-vivo detection method, device and electronic equipment | |
CN103604371A (en) | Mobile terminal and object measurement method thereof | |
US10878452B2 (en) | Method and system for dynamically adjusting displayed content based on analysis of viewer attributes | |
CN108324247B (en) | Method and system for evaluating skin wrinkles at specified positions | |
JP4159159B2 (en) | Advertising media evaluation device | |
KR20190088478A (en) | Engagement measurement system | |
US11576478B2 (en) | Method for simulating the rendering of a make-up product on a body area | |
CN102084397A (en) | Image processing device, method, and program | |
CN110111347B (en) | Image sign extraction method, device and storage medium | |
CN110826610A (en) | Method and system for intelligently detecting whether dressed clothes of personnel are standard | |
JP6957462B2 (en) | Hair curl measurement | |
CN110378304B (en) | Skin state detection method, device, equipment and storage medium | |
CN109145861B (en) | Emotion recognition device and method, head-mounted display equipment and storage medium | |
CN105872516A (en) | Method and device for obtaining parallax parameters of three-dimensional film source | |
CN108154142B (en) | Skin wrinkle evaluation method and system based on voice recognition | |
JP6527765B2 (en) | Wrinkle state analyzer and method | |
Patil et al. | Survey on approaches used for image quality assessment |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
C10 | Entry into substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant |